FLAP: a framework for linking free-text addresses to the Ordnance Survey Unique Property Reference Number database

Area of Science:

• Geospatial analysis
• Digital health
• Health informatics

Background:

• Linking free-text addresses to unique property identifiers is crucial for geospatial analysis in digital health.
• Existing tools lack open-source availability and validated performance for this task.
• Accurate address linking supports applications like identifying care home residents and optimizing resource allocation.

Purpose of the Study:

• To develop and validate a generalisable, open-source framework for linking free-text UK addresses to the Ordnance Survey Unique Property Reference Number (UPRN) database.
• To improve the accuracy and usability of address matching for digital health applications.
• To provide a machine learning-based solution adaptable to new data and continuous improvement.

Main Methods:

• Developed FLAP (A Framework for Linking free-text Addresses to Ordnance Survey UPRN database), a Python-based open-source tool.
• Employed a machine learning matching classifier combined with a fuzzy aligning algorithm for feature generation.
• Tested FLAP by linking Community Health Index (CHI) free-text addresses from NHS Tayside and NHS Fife to the UPRN database.

Main Results:

• Achieved an adjusted matching accuracy of 0.992 on a real-world test dataset.
• Demonstrated robustness against variations like typos, alternative formats, and incomplete address information.
• Enhanced usability through customizable matching thresholds and selection of top candidate records.

Conclusions:

• FLAP provides a high-performing and user-friendly solution for linking free-text UK addresses to the UPRN database.
• The framework's machine learning approach ensures adaptability and potential for ongoing improvement.
• FLAP facilitates essential geospatial analysis for digital health and resource management.